FocusNet: Imbalanced Large and Small Organ Segmentation with an End-to-End Deep Neural Network for Head and Neck CT Images

Gao, Yunhe; Huang, Rui; Chen, Ming; Wang, Zhe; Deng, Jincheng; Chen, Yuanyuan; Yang, Yiwei; Zhang, Jie; Tao, Chanjuan; Li, Hongsheng

doi:10.1007/978-3-030-32248-9_92

Cited by 65 publications

(61 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On training a single multi-class model to segment all the structures of interest, it was observed that the imbalance in class labels due to large differences in OAR sizes led to poor performance on smaller structures. Augmenting the loss function based on class frequencies did not produce significant improvement, as previously observed in [18]. Consequently, three separate models were developed-one to segment the chewing structures and one each for the larynx and the constrictor muscle.…”

Section: Methodsmentioning

confidence: 75%

“…We compared the performance of our method with H&N OARs segmentation models from the published literature. These included state-of-the-art CNN models such as the 3D Unet [13], FocusNet [18], a H&N OAR-focused CNN [12], commercial atlas-based segmentation tool SPICE [5] and other atlas-based methods [24,25]. The mean DSCs for the different methods are summarized in Table 4.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Deep learning-based auto-segmentation of swallowing and chewing structures

Iyer

Thor

Haq

et al. 2019

Preprint

View full text Add to dashboard Cite

Delineating swallowing and chewing structures in Head and Neck (H&N) CT scans is necessary during radiotherapy treatment (RT) planning to limit the incidence of speech dysfunction, dysphagia, and trismus. Automating this process is desirable so as to reduce the amount of manual labor required and to generate reproducible contours. However, this is a challenging problem due to the low soft tissue contrast and morphological complexity of the structures involved.In this work, we developed deep learning-based models using 194 H&N CT scans from our institution to automate the delineation of the masseters (left and right), medial pterygoids (left and right), larynx, and constrictor muscles using the DeepLabV3+ architecture. An ensemble of three models was developed for each group of structures using axial, coronal, and sagittal images to provide useful contextual information from the three different views. Each of these models was trained in 2.5D by populating the channels with three consecutive slices. The resulting probability maps were averaged, and voxels were assigned labels corresponding to the class with the highest combined probability. This method offers the advantage of providing increased redundancy in case of unintended failures of the individual models.A hold-out dataset of 24 H&N CT scans was used to evaluate the resulting models against expert contours in terms of Dice Similarity Coefficient (DSC) and 95 th percentile Hausdorff distance. Additionally, clinically relevant dose-volume histogram-based metrics were extracted from manual and auto-generated contours and compared to assess clinical utility. The models yielded DSC exceeding or comparable to the performance of previously reported autosegmentation methods. Differences in DVH metrics compared to ground-truth were not found to be statistically significant for all structures except the constrictor muscles. The resulting autosegmentation models are distributed through the open-source platform CERR, accessible at

show abstract

Section: Methodsmentioning

confidence: 75%

Section: Resultsmentioning

confidence: 99%

Deep learning-based auto-segmentation of swallowing and chewing structures

Iyer

Thor

Haq

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…The quantitative and qualitative evaluation results( Wang et al [35] Zhu et al [24] Zhu et al [24] (with additional dataset) Gao et al [26] proposed As the quantitative and qualitative evaluation results illustrated in Table V and Fig. 5, we can observe that: (1) the segmentation accuracies achieved by are inferior than and , especially for very large OARs(TLs) and small OARs(ONs, and chiasm).…”

Section: Discussionmentioning

confidence: 90%

“…Furthermore, the compatibility of the proposed methods on very large and small OARS such as temporal lobe, pituitary, and chiasm was not considered. Gao, et al [26] proposed a FocusNet to balance large and small OARs segmentation. It achieved more accurate segmentation on small OARs with training OAR specific model, which is timeconsuming.…”

Section: Introductionmentioning

confidence: 99%

“…The statistical analysis is implemented in SPSS 19.0 software in this study. The significant difference was defined by p<0.05.To compare the proposed method with other state of art methods, is also compared with five published methods[24,26,[33][34][35] on the MICCAI 2015 H&N OARs segmentation dataset[33], denoted as MICCAI'15 dataset (http://www.imagenglab.com/newsite/pddca/). The MICCAI'15 dataset consists of 38 samples for training and 10 samples for testing.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Prior Attention Enhanced Convolutional Neural Network Based Automatic Segmentation of Organs at Risk for Head and Neck Cancer Radiotherapy

et al. 2020

View full text Add to dashboard Cite

Aimed to automate the segmentation of organs at risk (OARs) in head and neck (H&N) cancer radiotherapy, we develop a novel Prior Attention enhanced convolutional neural Network (PANet) based Stepwise Refinement Segmentation Framework (SRSF) on full-size computed tomography (CT) images. The SRSF is built with a multiscale segmentation concept, in which OARs are segmented from coarse to fine. PANet is a pyramidal architecture with elements of inception block and prior attention. In this study, the developed PANet based SRSF is applied for OARs segmentation in H&N radiotherapy. 139 CT series and manually delineated contours of twenty-two OARs by experienced oncologists are collected from 139 H&N patients for training and evaluating the proposed PANet based SRSF. The mean testing Dice similarity coefficients (DSC) on 39 CT series range from 76.1±8.3% (left middle ear) to 91.9±1.4% (right mandible) for large volume OARs(mean volume>1cc) while the corresponding ranges are 63.4±12.3%(chiasm) to 81.0±14.1% (right lens) for small and challenging OARs(mean volume1cc). Furthermore, the proposed method also achieved superior segmentations over reference methods on the MICCAI 2015 H&N dataset with mean DSC of 95.6±0.7%, 81.3±4.0%, 77.6±4.5%, 77.5±4.6%, and 69.2±7.6%, on the mandible, left submandibular, left and right optical nerve, and chiasm, respectively. The accurate segmentation of OARs is obtained on both the self-collected testing data and public testing dataset, which implies that the proposed method can be used as a practicable and efficient tool for automated OARs contouring in the H&N cancer radiotherapy.

show abstract

Automatic segmentation of organs‐at‐risks of nasopharynx cancer and lung cancer by cross‐layer attention fusion network with TELD‐Loss

et al. 2021

View full text Add to dashboard Cite

Purpose: Radiotherapy is one of the main treatments of nasopharyngeal cancer (NPC) and lung cancer. Accurate segmentation of organs at risks (OARs) in CT images is a key step in radiotherapy planning for NPC and lung cancer. However, the segmentation of OARs is influenced by the highly imbalanced size of organs, which often results in very poor segmentation results for small and difficult-to-segment organs. In addition, the complex morphological changes and fuzzy boundaries of OARs also pose great challenges to the segmentation task. In this paper, we propose a cross-layer attention fusion network (CLAF-CNN) to solve the problem of accurately segmenting OARs. Methods: In CLAF-CNN, we integrate the spatial attention maps of the adjacent spatial attention modules to make the segmentation targets more accurately focused, so that the network can capture more target-related features. In this way, the spatial attention modules in the network can be learned and optimized together. In addition, we introduce a new Top-K exponential logarithmic Dice loss (TELD-Loss) to solve the imbalance problem in OAR segmentation. The TELD-Loss further introduces a Top-K optimization mechanism based on Dice loss and exponential logarithmic loss, which makes the network pay more attention to small organs and difficult-to-segment organs, so as to enhance the overall performance of the segmentation model. Results: We validated our framework on the OAR segmentation datasets of the head and neck and lung CT images in the StructSeg 2019 challenge. Experiments show that the CLAF-CNN outperforms the state-of -the-art attentionbased segmentation methods in the OAR segmentation task with average Dice coefficient of 79.65% for head and neck OARs and 88.39% for lung OARs. Conclusions: This work provides a new network named CLAF-CNN which contains cross-layer spatial attention map fusion architecture and TELD-Loss for OAR segmentation. Results demonstrated that the proposed method could obtain accurate segmentation results for OARs, which has a potential of improving the efficiency of radiotherapy planning for nasopharynx cancer and lung cancer.

show abstract

FocusNet: Imbalanced Large and Small Organ Segmentation with an End-to-End Deep Neural Network for Head and Neck CT Images

Cited by 65 publications

References 13 publications

Deep learning-based auto-segmentation of swallowing and chewing structures

Deep learning-based auto-segmentation of swallowing and chewing structures

Prior Attention Enhanced Convolutional Neural Network Based Automatic Segmentation of Organs at Risk for Head and Neck Cancer Radiotherapy

Automatic segmentation of organs‐at‐risks of nasopharynx cancer and lung cancer by cross‐layer attention fusion network with TELD‐Loss

Contact Info

Product

Resources

About